Investigation report: Neuronal Sensitization Drives Bone Regeneration via Energy Metabolism and BMP2-ErbB1 Signaling (Dai et al., Adv. Funct. Mater., DOI: 10.1002/adfm.76607)

Academic fraud report · Geng Detector

Summary

This investigation report assesses five concerns raised against the paper 'Neuronal Sensitization Drives Bone Regeneration via Energy Metabolism and BMP2-ErbB1 Signaling' by Qiyuan Dai, Zetao Wang, Zilin Li, Xiaodong Cao et al., published in Advanced Functional Materials (DOI: 10.1002/adfm.76607). The overall verdict, as articulated in the original report, is 'confirmed fabrication' (实锤). Three findings are treated as confirmed: a ZDOCK score of 13553.6 for the BMP2–ErbB1 pair that vastly exceeds the algorithm's plausible scoring range; an identical binding energy of −17.7 kcal/mol reported for two receptor pairs with very different ZDOCK scores; and use of a Malvern Mastersizer 3000 (a laser-diffraction particle sizer for sub-micron-to-millimeter materials) to characterize 100–200 nm sEVs/exosomes, which is methodologically inappropriate. Two further concerns (sample-size and timepoint inconsistencies between Figure 7 and the methods, and an extraneous space in a primer sequence in Table 4) are noted at lower severity. The report carries clear caveats: findings are AI-assisted, not a formal investigation outcome, and falsification must be confirmed by the institution or publisher.

Verdict

As stated in the original report: 🔴 实锤 ("confirmed fabrication"). Multiple findings point to apparent fabrication or severe methodological incoherence. A formal conclusion of misconduct requires investigation by the institution or publisher.

Key findings

Finding 1 (🔴, confirmed): Table 1 reports a BMP2–ErbB1 ZDOCK score of 13553.6, whereas other BMP2–receptor pairs (BMPR1A, BMPR1B, BMPR2, ErbB2, etc.) fall in the plausible range of roughly 1500–1900. ZDOCK, a rigid-body docking method based on 3D grid and pairwise-potential scoring, normally yields scores on the order of ~1000–3000; a value >13000 is reported as physically and mathematically outside the algorithm's range and is suspected to be manually fabricated or altered.
Finding 2 (🔴, confirmed): In Table 1, the binding energy of BMP2–ErbB1 is −17.7 kcal/mol and that of BMP2–ErbB2 is also exactly −17.7 kcal/mol, despite ErbB2 having a ZDOCK score of 1937.10—an order of magnitude below ErbB1. Identical binding energies for receptor pairs with very different sequences, conformations, and docking interfaces are statistically implausible.
Finding 3 (🔴, confirmed): Section 5.14 states that sEV particle size was measured with a Mastersizer 3000 (Malvern). Mastersizer 3000 is a laser-diffraction particle sizer (typical range ~0.1 µm to 3500 µm, i.e., 100 nm to 3.5 mm) intended for powders, soils, coatings, etc. Figure 6B reports sEVs of ~100–200 nm, well below the instrument's reliable range. Nanoparticle tracking analysis (NTA) or a Malvern Zetasizer Nano is the standard tool for exosomes.
Finding 4 (🟠, evidence-insufficient): Figure 7's in vivo bone-defect experiment is labeled n = 6, while Section 5.15 specifies euthanasia at 1 and 3 months post-op and Section 5.13 lists four independent biological replicates (n = 4) for RNA-seq. The in vivo design implies ≥12 rabbits plus potential attrition, expanding opportunities for data manipulation (p-hacking).
Finding 5 (🟡, confirmed): Table 4 lists the TrkA reverse primer as CACCGAGACCCC AAAAGGTG, containing a stray space. Primer sequences are continuous ATCG strings; this indicates careless copying from another source.

Evidence highlights

DOI: 10.1002/adfm.76607
Table 1 (p. 9): BMP2–ErbB1 ZDOCK = 13553.6; BMP2–ErbB2 ZDOCK = 1937.10; binding energies for BMP2–ErbB1 and BMP2–ErbB2 both equal to −17.7 kcal/mol.
Section 5.14 (p. 16): "The particle size distribution of sEVs was measured using typical DLS methods with Mastersizer 3000 (Malvern, America)" used for particles reported as 100–200 nm in Figure 6B.
Table 4 (p. 16): TrkA reverse primer spacing anomaly.
Figure 7 caption: n = 6 vs. Section 5.13 RNA-seq n = 4; Section 5.15 dual timepoints (1 and 3 months).

Notes

Confidence: High for Findings 1, 2, 3, and 5 (internally contradictory or physically implausible numerics/instruments). Lower for Finding 4, which is a workflow-scale concern rather than a smoking gun.
Limits of this report: Findings originate from a publicly available translation of a third-party AI-assisted review; the authors' rebuttal, raw ZDOCK log files, raw NTA measurements, and the original RNA-seq FASTQ files were not examined. Independent re-mining and institutional verification are required before any formal finding of misconduct.
Suggested follow-up: request authors' raw sEV-sizing data on a Zetasizer/NTA instrument; request raw ZDOCK parameters and logs; cross-check FASTQ files and metadata for the RNA-seq replicates; verify primer sequences against an order confirmation; consider formal PubPeer and editorial inquiries.